Regulation of photoreceptor gene expression by Crx-associated transcription factor network

Abstract

Rod and cone photoreceptors in the mammalian retina are special types of neurons that are responsible for phototransduction, the first step of vision. Development and maintenance of photoreceptors require precisely regulated gene expression. This regulation is mediated by a network of photoreceptor transcription factors centered on Crx, an Otx-like homeodomain transcription factor. The cell type (subtype) specificity of this network is governed by factors that are preferentially expressed by rods or cones or both, including the rod-determining factors neural retina leucine zipper protein (Nrl) and the orphan nuclear receptor Nr2e3; and cone-determining factors, mostly nuclear receptor family members. The best-documented of these include thyroid hormone receptor β2 (Trβ2), retinoid related orphan receptor Rorβ, and retinoid X receptor Rxrγ. The appropriate function of this network also depends on general transcription factors and co-factors that are ubiquitously expressed, such as the Sp zinc finger transcription factors and STAGA coactivator complexes. These cell type-specific and general transcription regulators form complex interactomes; mutations that interfere with any of the interactions can cause photoreceptor development defects or degeneration. In this manuscript, we review recent progress on the roles of various photoreceptor transcription factors and interactions in photoreceptor subtype development. We also provide evidence of auto-, para-, and feedback regulation among these factors at the transcriptional level. These protein-protein and protein-promoter interactions provide precision and specificity in controlling photoreceptor subtype-specific gene expression, development and survival. Understanding these interactions may provide insights to more effective therapeutic interventions for photoreceptor diseases.

1. INTRODUCTION

Photoreceptors in the vertebrate retina carry out phototransduction, the conversion of light into a neuronal signal that initiates the visual process. In rodents, about 73% of retinal neurons are photoreceptor cells (Young, 1985). Rods and cones, the two types of photoreceptors in the retina, show a species-specific ratio and spatial distribution. Rods are responsible for vision in dim light, while cones are responsible for color vision in bright light. Both rods and cones have unique cellular structures called outer segments, containing highly compact membrane discs where the visual pigment opsins and other machinery involved in phototransduction are densely packed. At the molecular level, photoreceptor cells preferentially express a set of genes that are essential for their function, so called photoreceptor-specific genes. Mutations in many of the photoreceptor specific genes are known to cause retinal degeneration diseases in humans. [For reviews, see (Hartong et al., 2006); and Retnet: http://www.sph.uth.tmc.edu/Retnet/]. Furthermore, the expression levels of these photoreceptor genes need to be precisely regulated. Increased or decreased expression levels of a wild-type photoreceptor gene can also lead to photoreceptor degeneration (Olsson et al., 1992; Humphries et al., 1997).

Precisely regulated photoreceptor gene expression is also a driving force for photoreceptor development/differentiation. Lineage tracing and birth dating experiments demonstrated that all of the neuronal cell types in the retina are derived from a common multi-potent progenitor cell (Turner and Cepko, 1987; Wetts and Fraser, 1988). For photoreceptor cells, cones are usually born (exit from the mitotic cycle and commit to the photoreceptor lineage) earlier than rods. In rodents, cones are born on embryonic days E11.5-E18.5, while rods are born in a longer period from E12.5 to postnatal day 7 (P7) with a peak at P0 (Carter-Dawson and LaVail, 1979; Young, 1985). However, there is a significant delay, several days in rodents, for newly born photoreceptor precursors to begin expressing the specific type of opsin and other genes that confer the mature phenotype (Watanabe and Raff, 1990; Cepko, 1996). During this lag time, the photoreceptor precursors appear to be “plastic” and can be induced to differentiate into different photoreceptor subtypes, depending on intrinsic and extrinsic regulatory factors (Nishida et al., 2003; Cheng et al., 2006; MacLaren et al., 2006; Roberts et al., 2006). The intrinsic factors mainly consist of transcription factors of homeodomain, bZIP and nuclear receptor families. In this manuscript, we review the recent progress in understanding these photoreceptor transcription factors, provide some evidence for the presence of network interactions among the major players, and present a model of how these interactions determine photoreceptor gene expression and development.

Photoreceptor transcription factors are the transcription regulators preferentially expressed by post-mitotic photoreceptor precursors and/or mature photoreceptors. Table 1 lists the major factors that are known to be important for photoreceptor development and maintenance, mostly based on in vivo loss-of-function studies. It is well established that members of bHLH and homeodomain transcription factor families play important roles in specifying various neuronal cell types in the retina [for review, see (Hatakeyama and Kageyama, 2004; Yan et al., 2005)], including photoreceptor precursors. Recently, though, much progress has been made in elucidating the roles of members of the nuclear receptor, bZIP and homeodomain families of transcription factors in specifying rod and cone photoreceptor subtypes. Below we will focus on these new findings and discuss some key factors in detail.

Table 1.

Transcription regulators for photoreceptor gene expression, development and/or maintenance

Factors	Expression in photoreceptors		Function2 / targets3	Key References
	Subtype	Period1
Photoreceptor-enriched
Homeodomain
Otx2	dev ph precursors	E10.5-P6	act / Crx, Rbp3, etc.	Nishida et al., 2003
Crx	dev/ad rods/cones	E12.5-ad	act / opsins, reg. factors	Chen et al., 1997; Furukawa et al., 1997; 1999
Rax	dev/ad ph	P10.5-ad	act / Rho, Arr, Rbp3	Zhang et al., 2000; Kimura et al., 2000
Qrx/RaxL	dev/ad ph	ad(Bo/Hu/Ch/Xe)	act / Rho	Chen et al., 2002; Wang et al., 2004
bZIP
Nrl	dev/ad rods	E13.5-ad	act / Rho, Nr2e3, Pde6b, etc.	Liu et al., 1996; Mears et al., 2001
bHLH
NeuroD1	dev rods	E12.5-P3	ph survival/unclear	Morrow, et al., 1999, Pennesi et al., 2003
Mash1	dev rods	E13.5-P3	act / Rho	Guillemot et al., 1993; Ahmad et al., 1995
Math5	dev precursors	E11.5-birth	rep / NeuroD1, Neurog2	Brown, et al, 1998; Le et al., 2006
Neurog2	dev precursors	E10-birth	act? / NeuroD1	Ma and Wang, 2006; Yan et al., 2005
Nuclear Receptors
Trβ2	dev cones	E16-ad	dual / S- and M-opsin	Yanagi et al., 2002; Roberts et al., 2006
Rxrγ	dev cones	E 14.5-ad	rep / S-opsin	Roberts et al., 2005
Rorβ	dev cones	E12.5-ad	act / S-opsin	Chow et al., 1998; Srinivas et al., 2006
Nr1d1(Rev-erb-α)	dev ph	P0-ad	dual? / circadian genes	Cheng et al., 2004
Nr2e1(Tlx)	dev cones	turned on E8	dual / S-opsin, Pax2, Rar	Zhang et al., 2006
Nr2e3	dev/ad rods	E18-ad	dual / all opsins	Kobayashi et al., 1999; Peng et al., 2005
Ubiquitously-expressed
Rb1	rods	P12-P21	act? / Nrl	Zhang et al., 2004
Sp1, Sp4	dev/ad ph	E12.5-ad	act / Pde6b, Rho	Lerner et al., 2001
ataxin-7	dev/ad ph	E12.5-ad	coact /opsins	Palhan et al., 2005
Gcn5	dev/ad ph	E12.5-ad	coact /opsins	Palhan et al., 2005
Cbp/p300	dev/ad ph	E12.5-ad	coact /opsins	Yanagi et al., 2000

Abbreviations: act-activator; ad-adult; coact-co-activator; dev-developmental; dual-activator and repressor; ph-photoreceptors; reg.-regulatory; rep-repressor; Arr-rod arrestin; Neurog2-neurogenin 2.

¹In mice or in species as noted: Bo-bovine, Hu-human, Ch-chicken, Xe-Xenopus.

²Based on loss-of-function studies.

³direct targets if known (based on protein-DNA binding assays).

1.1. Factors specifying the photoreceptor lineage - Otx2 and Crx

The role of Otx homeodomain transcription factors in eye development originally came from studies of Drosophila orthodenticle (Otd), a paired-type homeodomain protein that is required for the formation of anterior brain, eye and antenna in the fly (Finkelstein and Boncinelli, 1994). Subsequent studies showed that Otd plays an essential role in Drosophila photoreceptor development (Vandendries et al., 1996) by regulating the expression of opsin genes (Tahayato et al., 2003). Mammals have three Otd orthologs, Otx1, Otx2 and Crx, which is equivalent to Otx5 in fish, amphibians and chick (Plouhinec et al., 2003). The function of these Otd orthologs has diverged over time with Crx dedicated specifically to the development and maintenance of retinal photoreceptors and pinealocytes in the pineal gland involved in circadian regulation (see below). In terms of the protein sequence, the three mammalian Otd homologs share 87−88% homology in the homeodomain near the N-terminus and high homology in several discrete regions in the C-terminal portion, including a glutamine-rich region and the Otx-tail [(Chen et al., 1997; Furukawa et al., 1997); Figure 1]. Their homeodomain belongs to the K50 (lysine at position 50) paired-like class, similar to that of Drosophila bicoid protein, which, based on structure and functional studies, prefers to bind to DNA motifs with TAATCC or TAAGCT sequences (Treisman et al., 1989; Furukawa et al., 1997; Baird-Titus et al., 2006). These sequence motifs are widely present in the promoter or enhancer regions of many photoreceptor genes, including the opsin genes (Chen and Zack, 1996; Furukawa et al., 1997; Yu et al., 2006).

Figure 1. Schematic diagram of photoreceptor-specific transcription factors.

The domain structures of photoreceptor-specific transcription factors discussed in this paper are presented in scale. Conserved domains for classifying families of transcription factors are indicated in black, other regions of homology conserved among different family members are indicated by stippling. Functional regions are indicated above the box representing the factor; sites of mutations discussed in the text are indicated by arrowheads below the box. N- and C-terminals are indicated, and the number below the C-terminal end indicates the number of amino acids in the human protein. HOMEO, homeodomain; b, basic domain; L Zipper, leucine zipper domain; Zn F, zinc finger domain.

1.1.1. Otx2 specifies photoreceptor lineage by regulating the expression of Crx and other photoreceptor genes

Otx2 is expressed in the forebrain and midbrain neuroepithelium during development, including the eye domain. During development and in adults, Otx2 is expressed in several eye tissues, including neural retina and retinal pigmented epithelium (RPE) (Bovolenta et al., 1997). Otx2 is known to be required for the development and maintenance of the RPE by regulating the expression of RPE-specific transcription factors and genes (Martinez-Morales et al., 2001; Martinez-Morales et al., 2003). In the neural retina, Otx2 expression is seen in post-mitotic neuroblast cells that have the potential to develop into various cell types, including ganglion cells, bipolar cells and photoreceptor cells (Bovolenta et al., 1997; Baas et al., 2000). Nishida et al. carried out a parallel in situ hybridization analysis of Otx2 and Crx mRNA expression in developing mouse retina (Nishida et al., 2003) and showed that Otx2 expression is initially seen at E11.5. Its expression increases at E12.5 together with an induction of Crx expression, coinciding with early cone development. At E17.5, Otx2 expression is highly intensified in the outer part of the neuroblastic layer, where a Crx expression zone is established. After birth, when Crx expression reaches a peak and rod maturation begins around P5−6, Otx2 expression is down-regulated in the presumptive photoreceptor cell layer but up-regulated in the inner nuclear layer where bipolar and Muller glia cells are developing. The Otx2 spatial and temporal expression patterns suggest that Otx2 could play an essential role in photoreceptor development.

Human genetic studies

The human OTX2 gene maps to 14q21-q22, in an interval associated with microphthalmia and pituitary insufficiency. Using a candidate gene approach, Ragge et al., (2005) identified eight heterozygous OTX2 mutations from 333 patients with ocular malformations. The ocular phenotypes of these patients vary from severe bilateral anophthalmia to unilateral microphthalmia with Leber's congenital amaurosis (LCA). In vitro biochemical analysis (Ragge et al., 2005; Chatelain et al., 2006) suggests that these OTX2 mutations are likely to cause disease by a loss-of-function (haplo-insufficiency) mechanism, as many of the mutations reduce the ability of OTX2 to bind to DNA and/or activate the target gene promoter RBP3 in transfected HeLa cells.

Animal studies

Homozygous Otx2 knockout in the mouse is embryonic lethal due to defects in gastrulation and lack of rostral brain (Acampora et al., 1995; Matsuo et al., 1995; Ang et al., 1996). The heterozygous Otx2 knockout mouse (Otx2^+/−) showed multiple ocular defects, including microphthalmia, hyperplastic retina and RPE, and lack of lens, cornea and iris (Matsuo et al., 1995). To understand the role of Otx2 in photoreceptor development, Nishida et al (2003) generated a conditional Otx2 knockout in developing photoreceptors using a Cre transgene under control of the mouse Crx promoter. This Otx2 deficiency converts developing photoreceptors into amacrine-like cells in the retina, and completely blocks the formation of pinealocytes in the pineal gland. Thus, Otx2 is required for photoreceptor cell fate determination and pineal gland development. Otx2's role in photoreceptor development is also demonstrated by Otx2 over-expression studies (Nishida et al., 2003). Forced Otx2 expression in P0 rat retina using a retroviral vector results in a significant increase in the number of rod photoreceptors at the expense of bipolar, amacrine and Muller glia cells, suggesting that Otx2 promotes photoreceptor cell fate (Nishida et al., 2003). Given the observation that Otx2 expression switches to bipolar cells after the peak of photoreceptor development, these results also suggest that Otx2 may be involved in bipolar development as well. Forced Otx2 expression in adult iris- and ciliary-derived “stem” cells of rat origin is sufficient to induce the differentiation of photoreceptor-like cells (Akagi et al., 2004), consistent with Otx2 having a key role in specifying photoreceptor cell fate.

Mechanisms of action

Otx2 target genes in the photoreceptors are being studied by microarray analysis in Dr. Furukawa's laboratory in Japan. Although the microarray results remain to been seen, two direct Otx2 target genes are known. One is Rbp3 (Bobola et al., 1999; Fong and Fong, 1999), a Crx-independent gene (Furukawa et al., 1999) expressed by both rods and cones, and the other is Crx. Crx expression is abolished in the Otx2 conditional knockout mouse retina (Nishida et al., 2003). Otx2 significantly enhances Crx promoter activity in transient cotransfection assays. Chromatin immunoprecipitation analysis showed that Otx2 binds to the promoter region of Crx in vivo, further supporting the Crx gene as a direct target of Otx2 (see Results, Figure 3). Otx2 has also been reported to bind to and auto-activate its own promoter (Martinez-Morales et al., 2003). Furthermore, we have shown that Otx2 also binds to the promoter/enhancer region of several other known Crx targets, including rod and cone opsins, in the presence or absence of Crx. In transiently transfected HEK293 cells, Otx2 is also able to activate rhodopsin and M-cone opsin promoter activity, although less potently than Crx (Peng and Chen, 2005). These results suggest that Otx2 acts by directly regulating the expression of the key transcription factor Crx and its target genes.

Figure 3. ChIP analysis demonstrating that network transcription factors bind to their own and each other's promoters.

Antibodies against Crx, Nrl, Nr2e3, Trβ2, and NeuroD1 were used to immunoprecipitate the bound chromatin fragments from wild-type (WT), Crx^−/−, Nrl^−/−, or Nr2e3^rd7/rd7 (“Nr2e3^−/−“) retinae. Primers specific to the promoter regions of the genes listed on the left [(Peng and Chen, 2005); Table 3] were used to detect the presence of the candidate promoter regions in the immunoprecipitates by PCR. A band indicates that the transcription factor recognized by the immunoprecipitating antibody is bound to the promoter region of the indicated regulator or target gene. Target genes examined include S-opsin (Sop), M-opsin (Mop), rhodopsin (Rho), and interphotoreceptor binding protein (Rbp3), all of which are expressed in photoreceptors. GluR6, which is expressed in bipolar cells but not photoreceptors, serves as a control for photoreceptor specificity. In addition, PCR reactions using primers against DNA sequences immediately 3’ of each gene gave no bands (data not shown), confirming regulatory region-specific binding. Control immunoprecipitates using purified non-specific rabbit or goat IgG yielded no specific promoter sequences from WT (second lanes from the right) or knockout (data not shown) mice. Samples of retina homogenates (“input”) from WT (far right lanes) and knockout (data not shown) mice serve as positive controls for PCR.

1.1.2. Crx directly regulates the expression of many photoreceptor genes

Crx was identified by three laboratories independently. Using RT-PCR with degenerate primers corresponding to the paired-like homeodomain, Furukawa et al. cloned a murine Crx gene, which shows a photoreceptor-specific expression pattern (Furukawa et al., 1997). Chen et al. reported cloning bovine Crx using a yeast one-hybrid assay with a rhodopsin promoter element, Ret4, as bait (Chen et al., 1997) and demonstrated that Crx can bind to three target sites in the rhodopsin promoter as well as targets in several other rod gene promoters. Furthermore, Crx acts as a transcription activator and synergizes with the bZIP transcription factor Nrl in activating rhodopsin-reporter gene expression, suggesting for the first time that a high level of rhodopsin expression requires the function of at least two photoreceptor transcription factors. Using in situ hybridization analysis, Crx expression was found in both rods and cones, and in their precursors, starting at embryonic day 12.5 in mouse, coinciding with cone cell birth. Expression peaks at P5, correlating with the onset of rod photoreceptor maturation when rod-specific gene expression is turned up. BrdU incorporation assays confirmed that Crx expressing cells are post-mitotic photoreceptor precursors derived from those cells that have just exited the cell cycle and express Otx2 but not Pax6 (Garelli et al., 2006). Crx is the earliest expressed photoreceptor marker in the retina. It is also expressed in pinealocytes in the pineal gland and regulates photoentrainment (Furukawa et al., 1999) and expression of genes involved in synthesizing the circadian hormone melatonin in mice (Li et al., 1998). Interestingly, using immunohistochemistry studies, we have also found that Crx is expressed in rod bipolar cells that co-stain with the bipolar cell marker PKCα in both mouse and human retinas [(Wang et al., 2002); and data not shown], suggesting a possible role of Crx in bipolar cell function.

Human genetic studies

The first piece of evidence for Crx's role in development and maintenance of photoreceptors came from genetic studies performed by Freund et al. (1997), who cloned the human CRX gene based on its homology in the homeodomain to another retinal homeodomain protein, Chx10. The human CRX gene maps to 19q13.3, within a cone-rod dystrophy (CORD2) locus. Subsequent genetic screens not only identified CRX mutations in autosomal dominant cone-rod dystrophy (Freund et al., 1997), but also in autosomal dominant retinitis pigmentosa (adRP) (Sohocki et al., 1998) and Leber's congenital amaurosis (LCA) (Freund et al., 1998; Rivolta et al., 2001b). Most CRX mutations are inherited in an autosomal dominant manner or occur de novo, particularly in LCA cases (Rivolta et al., 2001a). Many mutations are nucleotide insertions or deletions resulting in formation of a premature stop codon 3’ of the mutated sites, which produce C-terminal truncated forms of CRX. Others are missense mutations, several of which are located in the homeodomain [(Rivolta et al., 2001a); see Figure 1]. In vitro functional analysis demonstrated that many of the disease-linked mutations altered the ability of CRX to bind to DNA (homeodomain mutations) and/or activate transcription of the rhodopsin gene (Chen et al., 2002). Thus, CRX mutations may cause disease by impairing CRX-mediated transcriptional regulation of photoreceptor genes. However, in vitro biochemical studies have not found a clear correlation between disease severity and the degree of biochemical abnormality, and it is not clear why CRX mutations cause dominant disease.

Animal studies

The second piece of evidence for Crx function came from knockout mouse studies. Homozygous Crx knockout mice (Crx^−/−) are blind at birth without detectable photoreceptor function, resembling the phenotype of LCA. Their photoreceptors never develop the outer segments critical for phototransduction, and subsequently degenerate (Furukawa et al., 1999). Serial analysis of gene expression (SAGE) performed on Crx^−/− retinae before the onset of photoreceptor degeneration showed that 46% of photoreceptor-enriched genes are Crx-dependent (Blackshaw et al., 2001), particularly the opsin genes, providing convincing evidence that altered photoreceptor gene expression is a primary cause of the Crx deficient phenotype. Heterozygous Crx knockout (Crx^+/−) mice, on the other hand, have normal photoreceptor function at the ages of 3 months or older. However, a delay in development of photoreceptor function was detected by electroretinogram (ERG) measures, despite normal appearance of the retina at one month of age (Furukawa et al., 1999). No photoreceptor degeneration was observed in Crx^+/− mice, raising the possibility that human diseases associated with CRX mutations could involve a dominant-negative effect on the Crx regulatory pathway.

The third piece of evidence for Crx function came from ectopic expression studies. Forced expression of recombinant Crx in P0 rat retina using a retroviral vector (Furukawa et al., 1997) induces rod differentiation, although less potent than forced Otx2 expression. As observed with Otx2, forced Crx expression leads to a reduction in the number of amacrine and Muller glia cells. However, the number of bipolar cells is unchanged. These results suggest that, like Otx2, Crx is an important factor for photoreceptor cell fate determination. In stem cell studies, forced Crx expression in adult rat iris- and ciliary-derived cells is sufficient to induce the formation of rhodopsin-expressing cells as potently as Otx2 (Haruta et al., 2001; Akagi et al., 2004). Similar experiments with primate stem cells, however, require both Crx and NeuroD1 to induce the photoreceptor phenotype (Akagi et al., 2005). These findings suggest that interaction with other photoreceptor transcription factors is important for Crx function.

Mechanisms of action

Crx is a trans-activator for many photoreceptor genes, based on gene expression profile studies. Using chromatin immunoprecipitation assays, we have shown that Crx activates transcription by directly binding to the promoter and/or enhancer regions of the target genes in photoreceptor cells (Peng and Chen, 2005). However, in transient cell transfection assays with target promoter-luciferase reporters, Crx alone has only a moderate transactivating activity (2−5-fold enhancement), even with the rhodopsin promoter, a well-known Crx target (Chen et al., 1997). Thus, one mechanism for Crx to activate transcription is to interact with other transcription regulators. Functional interactions with numerous other proteins have been reported. These include photoreceptor-specific transcription factors [Nrl and Nr2e3, discussed below; Qrx (Wang et al., 2004)] and general transcription factors [Sp family members (Lerner et al., 2001); and nuclear receptor Ror isoforms (Srinivas et al., 2006)]. Crx also interacts with chromatin remodeling factors [ataxin-7 (La Spada et al., 2001), HMG I/Y (Chau et al., 2000), Baf (Wang et al., 2002)], the transcription coactivators Cbp and p300 (Yanagi et al., 2000), and the STAGA coactivator/chromatin remodeling complex (Palhan et al., 2005). STAGA is a highly conserved multi-protein complex present from yeast (SAGA) to man (Martinez et al., 2001). One key component of STAGA is the histone acetyl-transferase (HAT) Gcn5 that catalyzes acetylation of histones, a chromatin modification often associated with transcriptional activation (Martinez et al., 2001). Crx interacts with STAGA via ataxin-7, a 110-kD protein in the STAGA complex. Expansion of the poly-glutamine tract of ataxin-7 is associated with a dominant neurological disorder, spinocerebellar ataxia type 7 (SCA7), which features cone-rod dystrophy-like retinal degeneration similar to the pathology linked to CRX mutations. Animal model studies and in vitro functional analysis suggest that Crx is a STAGA-dependent transcription activator (La Spada et al., 2001; Chen et al., 2004; Palhan et al., 2005). A polyglutamine-expanded ataxin-7 disrupts Gcn5 HAT activity, resulting in hypoacetylation of histones on Crx target genes and transcription impairment in a SCA7 transgenic mouse model. Thus, one possible mechanism for Crx transcription activation is to promote chromatin remodeling by recruiting STAGA or other HAT-containing coactivators to its target genes. This possibility is further supported by the findings that Crx also interacts with two other coactivators with intrinsic HAT activity, Cbp and p300 [(Yanagi et al., 2000); and data not shown].

One question related to Crx function is why the Crx deficient retina develops photoreceptor cells in the first place if Crx is critical for the expression of many photoreceptor genes. A possible explanation is that another Otd/Otx family member plays a redundant role with Crx in specifying photoreceptor cell fate, therefore partially compensating for Crx function in Crx^−/− mice. Indeed, the closely-related family member Otx2 is expressed in developing photoreceptors in the retina and up-regulated in the Crx^−/− mouse retina [(Furukawa et al., 1999); Table 2]. Apparently, Otx2 and Crx have redundant but indispensable roles in photoreceptor development and maintenance. These roles might be contributed by protein-protein or protein-promoter interactions between the two factors.

Table 2.

P14 retinal mRNA levels relative to WT (qRT-PCR analysis)

Genes	Crx−/−	Mouse strains Nrl−/−	Nr2e3−/−
Regulators
Otx2	2.50 ±0.12*	1.00 ±0.12	1.00 ±0.12
Crx	0.05 ±0.06*	1.00 ±0.06	1.03 ±0.06
Nrl	0.67 ±0.06*	0.06 ±0.06*	1.00 ±0.06
Nr2e3	0.61 ±0.05*	0.04 ±0.15*	0.04 ±0.17*
Trβ2	0.46 ±0.15*	1.02 ±0.15	1.02 ±0.17
Rxrγ	4.32 ±0.10*	1.57 ±0.15*	1.32 ±0.10*
Rorβ	2.22 ±0.15*	1.78 ±0.06*	1.28 ±0.10*
NeuroD1	0.75 ±0.15*	1.01 ±0.15	1.00 ±0.15
Targets
Sop	0.09 ±0.06*	2.14 ±0.10*	1.23 ±0.10*
Mop	0.06 ±0.06*	1.17 ±0.06*	1.19 ±0.10*
Rho	0.12 ±0.06*	0.07 ±0.06*	0.84 ±0.06*
Rbp3	1.00 ±0.10	1.00 ±0.15	1.00 ±0.15

Results of quantitative RT-PCR are presented as the ratio of the knockout to wild-type expression level of each gene (see Experimental Procedures for calculations). The numbers represent the mean value ± standard deviation from three repeats.

^*P < 0.05 based on paired student t-test.

1.2. Factors for rod development - Nrl and Nr2e3

1.2.1. Nrl specifies rod lineage in photoreceptor precursors

The neural retina leucine zipper protein (Nrl) is a basic-leucine zipper (bZIP) transcription factor belonging to the Maf subfamily (Swaroop et al., 1992). The Nrl cDNA was originally cloned from an adult human retina library by subtractive hybridization (Swaroop et al., 1992). The recombinant Nrl protein was subsequently found to bind to and regulate the rhodopsin promoter via NRE, an AP-1 like element located in the proximal rhodopsin promoter region (Kumar et al., 1996; Rehemtulla et al., 1996). RT-PCR and in situ hybridization analysis demonstrated that Nrl is highly expressed in the retina, although expression was also detected in the developing brain and lens (Liu et al., 1996). In the neural retina, native Nrl transcripts were seen at E14.5 (Liu et al., 1996) and stay on through the developmental period and into adulthood. Lineage tracing using a GFP transgene driven by the Nrl promoter and BrdU pulse-chase experiments in mouse retina showed that Nrl mRNA can be detected as early as E12.5 in those cells just completing terminal mitosis (Akimoto et al., 2006). These Nrl⁺ cells subsequently develop into rod photoreceptors. In addition, Nrl is also highly expressed in the pineal gland of the brain (Akimoto et al., 2006), suggesting a role in pineal gland development. At the protein level, Nrl in the nuclear fraction of human and mouse retinal extracts consists of multiple differentially phosphorylated isoforms ranging from 29 to 35 kDa on SDS-PAGE/Western blots (Swain et al., 2001; Kanda et al., 2007). The function of the phosphorylated Nrl isoforms remains to be determined, but they are more prominent at the peak of rod differentiation and are altered by some human NRL mutations [(Kanda et al., 2007); see below]. Immunostaining of Nrl in the human and mouse retina showed that Nrl is localized in the nucleus of rods but not cones (Swain et al., 2001). These results suggest that Nrl plays a role in rod development and maintenance.

Human genetic studies

The role of Nrl in rod function was first demonstrated by human genetic studies. The human NRL gene maps to chromosome 14q11.2 (Farjo et al., 1997). Subsequent mutation analysis of a large pedigree with autosomal dominant retinitis pigmentosa identified a S50T missense mutation in NRL that cosegregates with the disease (Bessant et al., 1999). Although NRL mutations are rare, additional missense mutations linked to adRP have been identified, with hot spots at residues S50 and P51 (Martinez-Gimeno et al., 2001; DeAngelis et al., 2002; Nishiguchi et al., 2004). Some of these hot spot mutations result in mutant forms of NRL that demonstrate reduced phosphorylation but hyperactivity in activating the rhodopsin promoter with CRX in vitro (Bessant et al., 1999; Nishiguchi et al., 2004; Kanda et al., 2007). This suggests that these are gain-of-function mutations. It is notable that the function of both rods and cones are more severely affected in patients with heterozygous NRL mutations than in patients with the RHODOPSIN mutation P23H (DeAngelis et al., 2002), raising the possibility that NRL mutations could actively affect cone function in humans. Recessive NRL mutations were also found in patients suffering clumped pigmentary retinal degeneration (Nishiguchi et al., 2004) with loss of rod, but normal blue cone function. These mutations are likely to be loss-of-function mutations as shown by in vitro function studies (Nishiguchi et al., 2004).

Animal studies

The most direct evidence for Nrl function in rod fate determination comes from Nrl knockout mouse studies (Mears et al., 2001). Knockout of Nrl results in loss of rod photoreceptor function based on ERG measures. However, the function of cones, especially S-cones, is significantly enhanced compared to that in wild-type mice, resembling the phenotype of enhanced S-cone syndrome (ESCS) in humans and rd7 in mice, both associated with Nr2e3 mutations (see below). Cone function in Nrl knockout mice is preserved as late as 31 weeks (Mears et al., 2001). Single-cell electrophysiology showed responses driven by both S- and M-opsin in all cells tested from Nrl knockout mice (Nikonov et al., 2005), as seen for the wild-type cones that co-express both opsins [(Applebury et al., 2000; Nikonov et al., 2006); see cone subtypes and gradients section below]. Although the dorsal/ventral M/S-cone gradient is preserved in Nrl^−/− mice, more S-opsin sensitivity is seen in Nrl^−/− cones than in wild-type cones from comparable dorsal/ventral levels of the retina (Nikonov et al., 2006). Consistent with the above electrophysiological measures, the ultrastructural analysis showed that the Nrl^−/− retina has much shorter cone-like outer segments with disrupted morphology and cone-like nuclei with decondensed chromatin (Mears et al., 2001; Daniele et al., 2005). Whorls and rosettes are seen at early ages and thinning of the outer nuclear layer occurs later on (Mears et al., 2001), similar to retinal histopathology in rd7 mice. The inner neurons of the rod pathway, specifically rod bipolar cells, horizontal cells and amacrine cells, make connections with these “trans-differentiated” cones (Strettoi et al., 2004). At the molecular level, the Nrl^−/− retina has no detectable expression of rod-specific genes, but a much higher level of S-opsin and moderately increased M-opsin expression [(Mears et al., 2001); see Results and Table 2]. Thus, knockout of Nrl essentially turns a rod dominant retina into an all-cone (or rodless) retina based on morphological, physiological and molecular criteria, suggesting that Nrl is required for rod cell fate determination.

To determine if Nrl is sufficient to induce the rod phenotype, Oh et al. (2007) studied transgenic mice that express Nrl in all photoreceptor precursors under the control of a Crx promoter. This ectopic Nrl expression converts the retina of either wild-type (WT) or Nrl^−/− mice to an all-rod retina. The inner neurons that ordinarily contact cones, such as ON cone bipolar and horizontal cells, make connections with rods in these mice. No cone-specific gene products were detected by RT-PCR or immunohistochemistry (Oh et al., 2007). Thus, Nrl is sufficient to guide photoreceptor precursors to a rod lineage. However, conditional expression of Nrl later during S-cone differentiation, using the S-cone promoter on the Nrl^−/− background, generated hybrid cells that co-expressed both rhodopsin and S-opsin. Although lineage-tracing experiments showed that some of the S-cones might have switched to a rod lineage, no rod function was detected by ERG in these mice (Oh et al., 2007), suggesting that differentiated photoreceptors may have limited plasticity, but require appropriate transcriptional regulation for maintenance. Another possible explanation, however, is that Nrl represses expression of S-cone genes, resulting in inactivation of cells already committed to the S-cone lineage.

Mechanisms of action

Gene expression profile studies demonstrated that Nrl is required for the expression of many rod genes, including the photoreceptor nuclear receptor Nr2e3 (Mears et al., 2001) that is linked to enhanced S-cone syndrome (ESCS) (see below). Thus, the phenotype of Nrl^−/− mice is in part contributed by the loss of Nr2e3 expression. This was further demonstrated by mouse studies showing that ectopic expression of Nr2e3 in photoreceptor precursors of Nrl^−/− mice can transform developing cones to rod-like photoreceptors (Cheng et al., 2006). Microarray analyses of the developing and mature retina of wild-type and Nrl^−/− mice have identified clusters of Nrl target genes (Yoshida et al., 2004; Yu et al., 2004; Akimoto et al., 2006). These not only include genes coding for phototransduction and structural or membrane associated proteins as expected, but also transcriptional regulators, intracellular transport proteins and components of known signaling pathways. The Bmp/Smad pathway is repressed in the Nrl^−/− retina, as Bmp (Bmp2, 4 and Bmpr1a) and Smad (Smad 1, 5, and 4) genes are down regulated (Yu et al., 2004). Chromatin immunoprecipitation assays demonstrated that Nrl binds to the promoter of Bmp2 and Bmp4, suggesting they are direct targets of Nrl. In contrast, many components of the Wnt/Ca²⁺ signaling pathway showed altered (either up- or down-regulated) expression in Nrl^−/− retina. These results suggest that Nrl is also important for maintaining rod function and homeostasis by integrating various signaling pathways.

In the Nrl^−/− mouse, the expression of all rod-specific genes is lost but cone genes are upregulated, suggesting that Nrl represses cone genes either directly or indirectly. Chromatin immunoprecipitation assays showed that Nrl can directly bind to cone gene promoters, including S-opsin, M-opsin, cone arrestin (Arr3), and the cone transcription factor Trβ2 [(Peng and Chen, 2005; Oh et al., 2007) and Figure 3], suggesting that Nrl may directly regulate cone gene expression in rods. On the other hand, transient co-transfection assays with luciferase reporter constructs driven by either M-cone or S-cone opsin promoters showed that Nrl actually activates cone opsin promoters in vitro (Peng et al., 2005). It is known that the retinoid X receptor gamma (Rxrγ) that represses S-cone expression (see below) is upregulated in the Nrl^−/− mouse retina. Furthermore, the expression of cone genes is upregulated in rd7 mouse retina where Nr2e3 protein is missing, but Nrl is normally expressed (Peng et al., 2005). These findings suggest that Nrl itself may not act as a transcription repressor but rather indirectly repress cone genes via the function of Rxrγ, Nr2e3, or other transcription repressors (see below).

Nrl is known to interact with several transcriptional regulators, in addition to the possibility of forming heterodimers with other bZIP family members expressed in the retina, such as Jun/Fos family members (He et al., 1998). Nrl was the first protein identified that acts synergistically with Crx to activate the rhodopsin promoter (Chen et al., 1997; Mitton et al., 2000). This synergy, however, is not observed for cone opsin promoters (Peng and Chen, 2005). Nrl and Crx interact through the leucine zipper and homeodomain, respectively (Mitton et al., 2000). Two mutations in the CRX homeodomain identified in human patients (R41W and R90W) decreased this interaction (Mitton et al., 2000). Nrl also interacts with the TATA-binding protein, Tbp, through a different domain near the N-terminal that is conserved among Maf family members (Friedman et al., 2004). Thus, Nrl may function by recruiting or stabilizing Tbp, which then recruits the general transcription complex (Friedman et al., 2004). Nrl was also reported to interact with Fiz1, a zinc finger protein that complexes with Flt3 receptor tyrosine kinase (Mitton et al., 2003). Fiz1 potentiates Nrl or Crx/Nrl activity on rhodopsin and PDE6B promoters (Mali et al., 2007), further implicating Nrl's involvement in coordinating the intrinsic photoreceptor developmental program with extracellular signaling pathways.

1.2.2. Nr2e3 is a dual transcription regulator required for terminal differentiation of rods

The nuclear orphan receptor Nr2e3 (photoreceptor-specific nuclear receptor, PNR) was originally identified by its homology to the orphan nuclear receptor tailless (Tlx, Nr2e1) initially found in Drosophila, and by its specific expression in retinal photoreceptor cells (Kobayashi et al., 1999). Similar to other members of the Tlx nuclear receptor family, Nr2e3 has a zinc-finger DNA-binding domain near the N-terminus and a ligand-binding domain in the C-terminus for a ligand yet to be identified [(Kobayashi et al., 1999); Figure 1]. Nr2e3 is expressed in retinal photoreceptor cells beginning around E18 in the mouse, peaking during rod differentiation and persisting into adulthood at a decreased level (Kobayashi et al., 1999; Takezawa et al., 2007). Expression appears primarily localized to the outer nuclear layer (Kobayashi et al., 1999; Haider et al., 2000; Haider et al., 2001). This persistent expression, mainly in rods in mammals (Bumsted O'Brien et al., 2004; Chen et al., 2005; Peng et al., 2005), suggests that Nr2e3 plays a major role in rod differentiation and maintenance. On the other hand, some early cone precursors appear to transiently express Nr2e3, at least in lower vertebrates (Chen et al., 2005) and Nr2e3 expression has been reported in other retinal cell types (Chen et al., 1999) including mouse cones [(Haider et al., 2006); see below], so it may also be involved in other developmental processes.

Human genetic studies

Mutations in human NR2E3 cause enhanced S-cone syndrome (ESCS), an autosomal recessive disease featuring hyperfunction of blue cones, defective function of rods, and blindness in the late stages (Haider et al., 2000; Jacobson et al., 2004; Wright et al., 2004). Histopathological studies showed excess S-cones in the retina, some of which express both S- and M-cone opsins (Milam et al., 2002), which is unusual in humans (Lukats et al., 2005; Peichl, 2005).

Animal studies

The rd7 mouse, a naturally occurring mutant strain, resembles the phenotype of human ESCS. Genetic analysis revealed that these mice carry a homozygous 380-bp deletion in the coding region of the Nr2e3 cDNA (Akhmedov et al., 2000; Haider et al., 2001), due to mRNA splicing defects resulting from an L1-retrotransposon insertion (Chen et al., 2006). This leads to a frame-shift with a premature termination. Subsequent studies demonstrated that the rd7 mouse does not produce Nr2e3 protein, and therefore represents a bona fide null mutant of Nr2e3 (Nr2e3^−/−) (Peng et al., 2005; Haider et al., 2006). The rd7 mouse retina contains whorls and rosettes in the photoreceptor layer at early ages, followed by slow photoreceptor degeneration. Similar to ESCS in humans, the rd7 retina has excess cones that express mostly S-cone opsin (Haider et al., 2001). Rod and cone function measured by electroretinography (ERG) is near normal in young adults but significantly declines in older mice as a result of degeneration of both rods and cones (Akhmedov et al., 2000; Ueno et al., 2005; Haider et al., 2006). These phenotypes of human ESCS and rd7 mice support a role for Nr2e3 in rod/cone development by demonstrating how its mutations lead to disease.

Mechanisms of action

The phenotype of Nr2e3 mutants suggests two possible hypotheses for Nr2e3 function in vivo: 1) Nr2e3 mutations cause defects in cell-fate determination resulting in transdifferentiation of developing rods into cones, or 2) Nr2e3 mutations result in abnormal cone proliferation leading to an increase in the absolute number of photoreceptors as well as disrupting the rod/cone ratio. Several pieces of evidence strongly support the first hypothesis: a) As a downstream target of the rod specific transcription factor Nrl, Nr2e3 is predominantly expressed in post-mitotic developing and mature rods. In Nrl^−/− mice, enhanced S-cones arise from postmitotic rod precursors and Nr2e3 expression is completely abolished (Mears et al., 2001). b) Morphological and microarray analysis of the rd7 retina indicate that the majority of photoreceptors exhibit a hybrid rod-cone phenotype, i.e. expressing both rod and cone genes (Chen et al., 2005; Corbo and Cepko, 2005). c) Direct target gene studies suggest that Nr2e3 is a dual transcription regulator for both rod and cone genes. In vitro protein-DNA binding assays initially showed that Nr2e3 recognizes a consensus DNA sequence with a direct repeat and binds as a homodimer (Kobayashi et al., 1999; Chen et al., 2005). This binding appears to mediate transcriptional repression (Chen et al., 2005) rather than activation as previously observed for the rhodopsin promoter (Cheng et al., 2004). The consensus Nr2e3 DNA recognition sequence has actually not been found in native opsin promoters or other known target genes, although the half site is present (Peng et al., 2005). Chromatin immunoprecipitation assays subsequently demonstrated that the Nr2e3 protein is indeed associated with both rod and cone opsin gene promoters in the retina of wild-type and “coneless” mice (Peng et al., 2005; Peng and Chen, 2005). However, this association depends on the Crx protein, as it does not occur in the Crx^−/− retina in spite of the presence of Nr2e3 protein (Peng et al., 2005). Furthermore, in transient transfection assays, Nr2e3 alone has limited regulated activity on target opsin gene promoters. In the presence of Crx and Nrl, however, Nr2e3 potentiates Crx/Nrl activation of rhodopsin (Cheng et al., 2004; Peng et al., 2005), but represses their activity on the cone opsin promoters (Chen et al., 2005; Peng et al., 2005). In addition, four genetically identified NR2E3 missense mutations demonstrated altered dual regulatory activity (Peng et al., 2005). Consistent with these in vitro results, quantitative RT-PCR analysis demonstrated that the rd7 retina exhibits down-regulated rod gene expression, but up-regulated cone gene expression during photoreceptor differentiation (Peng et al., 2005). These suggest that Nr2e3 acts as a dual regulator to promote rod phenotype differentiation and suppress cone gene expression in developing rods by modulating Crx/Nrl activity on rod and cone promoters. d) Gain-of-function studies have shown that ectopic expression of Nr2e3 using a Crx promoter in Nrl^−/− retina is sufficient for guiding photoreceptor precursors to develop into rod-like cells that express rhodopsin but not cone opsins (Cheng et al., 2006). Ectopic expression of Nr2e3 driven by the S-opsin promoter is also sufficient to transform differentiating S-cones into rod-like cells. Altogether, these findings suggest that Nr2e3 acts downstream of Nrl and Crx to reinforce the development of the rod phenotype by suppressing the cone pathway.

These findings, however, do not necessarily rule out the second possibility that Nr2e3 has a function in limiting proliferation of early S-cone precursors (Haider et al., 2001; Yanagi et al., 2002). As seen in zebrafish retina (Chen et al., 2005), using more sensitive immunostaining assays with GFP-labeled cones, Nr2e3 was recently found to be expressed in developing and mature M/S-cones (Haider et al., 2006). Some Nr2e3 positive cells in E18 mouse retina also express Ki67, a mitotic marker, suggesting Nr2e3 is expressed in mitotic cells of the developing retina. BrdU incorporation assays in rd7 retina at late developmental stages, including P14 and P30, demonstrated prolonged proliferation of ectopic retinal progenitors that subsequently develop into S-cones. No S-cones were observed to re-enter the cell cycle, however. Increased apoptosis is also seen in late developing stages of rd7 retina. These results support a role of Nr2e3 in suppressing cone proliferation.

The molecular mechanisms by which Nr2e3 plays a dual regulatory role in rod vs. cone gene expression remain to be determined, but are expected to involve interactions with co-activators and co-repressors. Nr2e3 is known to interact with another nuclear receptor, Nr1d1 (Rev-erb-α) (Cheng et al., 2004), forming a complex with Crx/Nrl that potentiates rhodopsin promoter activity. Nr1d1 is a member of the circadian clock (Yin et al., 2006) involved in regulating diurnal variations in gene expression. It is also known that the expression of some photoreceptor genes such as rhodopsin is under circadian regulation (Bowes et al., 1988; von Schantz et al., 1999). Thus the Nr2e3/Nr1d1 interaction may play a role in regulating the diurnal expression pattern of these genes. Nr2e3 has also been reported to interact with Rarα and Rxrβ (Chen et al., 1999). As for possible co-repressors, its closely-related family member Nr2e1 (Tlx), was reported to interact with the co-repressor atrophin1 (Atn1) in the retina (Zhang et al., 2006), raising the possibility that Nr2e3 might also use Atn1 as a co-repressor. Interestingly, Nr2e1 plays a role in modulating retinal progenitor cell proliferation and cell cycle re-entry by inhibiting the expression of Pten, a negative regulator of neural stem cell proliferation (Zhang et al., 2006). An Nr2e1 null mutation also results in enhanced S-cone syndrome in mice (Zhang et al., 2006). Takezawa et al. (2007) identified a novel cell cycle-dependent Nr2e3 co-repressor named Ret-CoR, which is preferentially expressed in the developing retina and brain as well as Y79 retinoblastoma cells. Ret-CoR expression is down regulated in mature retina, but clearly present in the photoreceptor nuclear layer where Nr2e3 is expressed. As reported for the other nuclear co-repressors NCoR/SMRT, Ret-CoR forms a multiprotein complex containing histone deacetylases (HDAC 1/2 and 3), NCoR and Rb/p107 that are known to regulate cell cycle progression. Nr2e3 appears to recruit Ret-CoR to the promoter of Cyclin D1, which is required for proliferation of retinal progenitor cells, and repress its expression. This new finding supports a possible role of Nr2e3 in regulating cell proliferation as discussed above. Altogether, these findings suggest that Nr2e3 promotes rod differentiation by bi-directionally regulating rod vs. cone gene expression and possibly inhibiting the proliferation of developing cones.

1.3. Factors for cone development

1.3.1. Cone subtypes and gradients

Cones are less sensitive to light than rods, but they provide visual acuity, the ability to distinguish features of the visual environment. Consequently, their distribution across the retina is not random, but varies among species to reflect each species' visual needs [(Peichl, 2005); Figure 2]. Most vertebrates have at least two different cone subtypes, producing opsins with different spectral sensitivities. In addition, the cone outputs are subjected to more processing in the retina than rod signals in order to extract more visual information, so cones make synaptic connections with a larger number and greater variety of inner retina cells than rods. Conversely, in order to preserve spatial resolution, cone signals are not summed by converging on a small number of output cells the way rod signals are (Peichl, 2005). The result is that more inner retina and central visual pathway neurons are involved with processing cone signals than rod signals. This has implications for understanding retina development, and should also be taken into account in experiments in which photoreceptor fate is genetically manipulated.

Figure 2. Distribution of cones and opsin expression in mouse and human retina.

A. In the mouse, cones are scattered throughout the retina, with M-cones (green) predominating in the dorsal retina and S-cones (blue) predominating in the ventral retina. Many cones express both opsins. B. M-opsin and S-opsin are expressed in complementary dorsal (D) to ventral (V) gradients across the mouse retina, likely in response to a gradient of thyroid hormone (TH), which is established between P4 and P10, and is highest in the dorsal retina as indicated. C. This graph shows the spatial density of rods (blue) and cones (orange) in a horizontal strip of human retina across the fovea (centered at position 0) and optic disc. Cones are concentrated in the fovea, and rods are excluded from this region. Elsewhere in the retina, rods predominate and cones are sparse. D. Within the fovea, cones expressing either green or red opsin predominate, with cones expression blue opsin found sparsely around the peripheral fovea region. Human cones only express a single type of opsin.

Panels A and B are from (Applebury et al., 2000), reprinted by permission from Cell Press, with TH gradient added from (Roberts et al., 2006). Panel C is from (Rodieck, 1998), pg. 43, reprinted by permission from Sinauer Associates, Inc. Panel D is from (Cepko, 2000), Figure 2, reprinted by permission from Macmillan Publishers, Ltd: Nature 24: 99−100, copyright 2000.

Most studies aimed at understanding the genetic mechanisms underlying human cone development have been performed in mice, which have two subtypes of cones. Mice do not have a cone-rich area in their retina like primates do, but the two cone subtypes are distributed in inverse gradients across the retina (Figure 2A-B). The density of cones expressing short-wavelength sensitive S-opsin is highest in the ventral retina, and cones expressing longer wavelength sensitive M-opsin are most dense in the dorsal retina. In the region where these two gradients overlap, many cones express both photopigments (Rohlich et al., 1994; Applebury et al., 2000; Lukats et al., 2005). As in most vertebrates, there is a region of higher visual sensitivity along the equator of the retina that corresponds with a disproportionately increased area of representation in the visual cortex (Luo et al., 2006). In the mouse retina, S-cones differentiate first, beginning shortly after birth, with most differentiating cells localized to the ventral retina. M-cones begin differentiating about a week later, mostly in the dorsal retina and coinciding with a decrease in the number of cells expressing S-opsin (Cepko, 1996; Roberts et al., 2006). Increasing evidence suggests that the cone gradients are generated during development by the action of diffusible growth factors or hormones (extrinsic signals) and their receptors, mainly nuclear receptor family transcription factors (intrinsic program). These receptors exist as families of related genes that have apparently diverged from a common ancestral precursor (Germain et al., 2006). In the following section, we will focus on recent progress on the role of nuclear receptors in cone development and discuss the possible action of their ligands.

1.3.2. Factors that influence S-cone differentiation

Extrinsic signal – to be determined

The regulatory factors that are responsible for triggering cone progenitors to begin differentiating are still not well understood. Retinoic acid or related compounds are tempting candidates, since they have been shown to induce expression of photoreceptor specific genes and to influence cell fate choices in vitro (Hyatt and Dowling, 1997). It is also known that during early eye development in rodents, the equator region of the retina expresses retinoic acid degrading enzymes, while the rest of the retina expresses enzymes that convert Vitamin A to retinoic acid, thus establishing a discontinuous gradient of retinoic acid signaling across the retina (Luo et al., 2006). However, there is currently no compelling evidence that retinoic acid or related compounds influence cone photoreceptor distribution in vivo, although these studies led to interesting findings regarding the receptors for these compounds.

Rxrγ and Trβ2 are negative regulators for S-cones

Retinoic acid and related compounds work by diffusing through cellular membranes and binding to nuclear receptors, which associate with specific regulatory elements in the DNA of gene promoter and enhancer regions. These receptors exist as families of related genes. The retinoic acid receptor (Rar) and retinoid-related receptor (Rxr) families both consist of three genes, A, B, and C, producing the α, β, and γ isoforms, respectively, of the receptors. Extensive knockout mouse studies have been performed to try to determine which isoforms are expressed in particular tissues and mediate signals for particular processes or sets of genes. Most isoforms are expressed in the retina (Janssen et al., 1999; Mori et al., 2001), but the Rxrγ receptor is of particular interest because it is localized to developing cone photoreceptors in a number of species. Knockout of Rxrγ in mice destroys the gradient of S-cone distribution and results in S-opsin expression in all cones in the retina (Roberts et al., 2005), indicating that its likely role in S-cones is inhibitory rather than inductive. Rxrs are unique among nuclear receptors because they heterodimerize with members of several other nuclear receptor families, including Nr2e3 (Chen et al., 1999) and thyroid hormone receptors (Szanto et al., 2004). Thyroid hormone receptor Trβ2 is likely to be the heterodimerization partner involved in suppressing S-cones (see below).

Rorβ2 is a positive regulator for S-cones

Retinoid-related orphan receptors (Ror) are another family of genetically-related receptors with homology to the receptors for retinoids, but whose actual ligands have not been identified. One isoform in particular, Rorβ2, is expressed in photoreceptors as well as other cells in regions of the brain involved in regulating circadian rhythm (Andre et al., 1998). This isoform is expressed early in retinal progenitor cells (beginning at E13.5 in rat) and appears to increase their proliferation (Chow et al., 1998). The other product of this gene, Rorβ1, is not produced in the retina (Azadi et al., 2002). As development proceeds, expression of Rorβ2 continues at a lower level in photoreceptor cells, as well as in amacrine and ganglion cells (Chow et al., 1998). Rorβ2 has also been shown to synergize with Crx in vitro to activate the S-opsin gene (Srinivas et al., 2006). Mice that are homozygous for knockout of Rorβ fail to express S-opsin at the appropriate developmental time, although M-opsin expression is unaffected. However, in these mice the outer nuclear layer is disorganized and photoreceptor outer and inner segments are not produced, indicating that Rorβ2 may have several additional roles in differentiation of both rod and cone photoreceptors (Srinivas et al., 2006). Yanagi et al. (2002) suggest that the S-cone phenotype is a default state, to explain why disruption of either rod-inducing factors (Nrl or Nr2e3) or factors involved in cone differentiation (Rxrγ or Trβ2) result in over-production of S-cones. The discovery of this positive regulator of S-opsin, however, suggests that the S-cone phenotype might be actively selected, arguing against the hypothesis of a default pathway.

1.3.3. Factors that influence M-cone differentiation

Extrinsic signal – thyroid hormone

M-cones develop later than S-cones in rodents, but the mechanisms involved are currently more completely understood. The M-cone gradient (Figure 2A-B) is most likely established by thyroid hormone (TH). This hormone is produced by the thyroid gland and distributed by blood circulation to the tissues, where it is partially deiodinated to the active form (3,5,3’ tri-iodothyronine, or T3) (Forrest et al., 2002). Its involvement in mediating the visual changes that occur during amphibian metamorphosis have long been known (Hoskins, 1990), and it was also shown to influence chick and rat retinal progenitor development. Harpavat and Cepko reviewed the role of TH in retinal development in 2003 (Harpavat and Cepko, 2003). TH is distributed uniformly across the retina at birth, but during the period of M-cone differentiation between P4 and P10 forms a gradient with higher concentrations in the dorsal than ventral retina [(Roberts et al., 2006); Figure 2B]. This implicates TH as the extrinsic signal responsible for establishing the M-cone gradient.

Trβ2 is a positive regulator for M-cones

Vertebrates have two genes for TH receptors, and each produces several protein isoforms [reviewed in (Forrest et al., 2002; Eckey et al., 2003)]. The thyroid hormone receptor Trβ2, a splice variant of the thyroid hormone receptor B (Thrb) gene, is implicated in photoreceptor development in chick and mouse, based on its expression in retinal progenitor cells and developing photoreceptors. In mouse eyes, expression of Trβ2 begins about E16, peaks around E18 as cone photoreceptors begin differentiating, then decreases (Ng et al., 2001; Yanagi et al., 2002), but the expression is uniform across the retina (Ng et al., 2001; Roberts et al., 2005; Roberts et al., 2006). However, during the time M-cones are developing, its ligand TH becomes distributed in a gradient with higher concentrations in the dorsal than ventral retina [(Roberts et al., 2006); Figure 2B].

In vitro and animal studies

Transient transfection studies showed that Trβ2, in the presence of TH, activated M-opsin transcription and inhibited Crx-mediated transcription of S-opsin (Yanagi et al., 2002). Addition of exogenous TH also promoted M-opsin expression and inhibited S-opsin in embryonic retina explant cultures from wild-type (WT) but not Trβ2 knockout mice (Roberts et al., 2006). Furthermore, daily injection of TH into WT mouse pups beginning on P0 drastically decreased the number of S-cones found in all parts of the retina three days later. No decrease in S-cone numbers was seen in Trβ2 knockout mice treated similarly (Roberts et al., 2006). These results showed that Trβ2 induced M-opsin expression and concurrently inhibited S-cone production in a ligand-dependent manner. Knockout of the photoreceptor-specific Trβ2 isoform of the Thrb gene converted all cones to the S-phenotype, resulting in loss of both M-opsin expression and the S-cone gradient in vivo (Ng et al., 2001). This phenotype is also reproduced in a mouse with a knockin mutation in the DNA binding domain of Thrb that abolishes specific DNA sequence binding without affecting ligand binding or cofactor interactions (Shibusawa et al., 2003). These findings indicate that both M-opsin induction and the establishment of the S-cone gradient depend on Trβ2 DNA binding. The role of Trβ2 thus appears to be to induce a subset of developing cones to further specialize as M-cones (Yanagi et al., 2002) by suppressing expression of S-opsin (and possibly other S-cone genes) but promoting expression of M-opsin (and possibly other M-cone genes).

Mechanisms of action

TH nuclear receptors are reported to exert their effects as heterodimers in combination with retinoid (usually Rxr) receptors (Mangelsdorf and Evans, 1995). The Rxrγ receptor mentioned previously is a likely candidate for the Trβ2 dimerization partner in developing cones. Rxrγ itself is not involved in M-opsin expression, however, since the M-cone gradient forms normally in Rxrγ knockout mice (Roberts et al., 2005). Roberts et al. (Roberts et al., 2006) hypothesize that a Trβ2:Rxrγ heterodimer is responsible for suppressing S-opsin expression, while binding of TH induces dissociation of Rxrγ and formation of Trβ2 homodimers, which activate M-opsin expression. This would also explain the inhibition of S-opsin expression in the dorsal retina prior to P6. Exogenous TH could overcome this inhibition if it induced dissociation of receptor heterodimers prematurely, before the gradient of thyroid hormone becomes established. Thyroid hormone receptors mediate repression (usually in the absence of ligand) by interacting with nuclear co-repressors NCoR and/or SMRT (Eckey et al., 2003; Makowski et al., 2003; Havis et al., 2006). In fact, a growing body of evidence indicates that heterodimers of thyroid hormone and Rxr receptors adopt different configurations based on information provided by the DNA binding site with which they associate, that facilitate or block interactions with co-activators or co-repressors [(Harvey et al., 2007) and references cited therein; (Ghosh et al., 2002; Diallo et al., 2007)]. Thus, the action of a nuclear receptor heterodimer can be influenced by each target promoter sequence to fine-tune interactions with ligands and cofactors. Since S-cones develop several days before M-cones appear in mice but are less prevalent in the dorsal retina, a repressive mechanism must exist for suppressing S-cones (or at least expression of S-cone genes) in regions where M-cones will predominate. Increasing evidence therefore supports a dual role for Trβ2, in conjunction with Rxrγ, in mediating this repression.

The actual role of Trβ2 on S-opsin expression may be more complex than the simple inhibition implied above. Findings from a study designed to map expression quantitative trait loci (eQTL) in the rat showed that point mutations affecting a conserved serine and another residue in the N-terminal domain of Trβ2 (Ser56Asn and His58Arg) correlated with decreases in S-opsin expression levels of as much as 30% in homozygotes (Scheetz et al., 2006). The affected residues fall within a poorly characterized ligand-independent transactivation domain that can interact with cofactors and is the target of post-translational modification in some nuclear receptors (Germain et al., 2006). Trβ2 binds directly to the S-opsin promoter (Figure 3) and has been reported to interact with the basal transcription machinery as well as co-activators and co-repressors to exert complex regulatory effects on target genes (Eckey et al., 2003). The mutations identified by Scheetz et al. (2006) might therefore alter one or more of these interactions, making activation of the S-opsin gene less efficient in the presence of the mutated receptor.

2. RESULTS AND DISCUSSION

Photoreceptor transcription factors form a network to regulate the expression of themselves and each other

During the course of reviewing and studying the above rod and cone transcription factors, we have observed evidence of cross-talk and feedback regulation among these factors at the transcriptional level. We hypothesized that each of these factors regulates its own expression and that of the other factors by direct binding of the transcription factor protein to promoter elements in the DNA. Such interaction and feedback regulation could be important for regulating development and maintenance of each of the photoreceptor phenotypes. To test this hypothesis, we performed chromatin immunoprecipitation (ChIP) and quantitative RT-PCR analysis of five of these transcription factors: the photoreceptor lineage determinant Crx, rod-lineage determinants Nrl and Nr2e3, the cone determination factor Trβ2, and the HLH factor NeuroD1 that has been shown to be important for photoreceptor survival (Morrow et al., 1999; Pennesi et al., 2003). Chromatin immunoprecipitation (ChIP) assays were performed on the retinas of wild-type, Crx^−/−, Nrl^−/− and Nr2e3^rd7/rd7 (labeled as “Nr2e3^−/−“) mice at the age of P14 when these factors are all expressed but before photoreceptor degeneration begins in the mutant mice. The ChIP results were analyzed using PCR with primers spanning the promoter region of each regulatory factor; their dependent target genes, rod and cone opsins; and their independent gene Rbp3 (Figure 3). PCR assays with primers spanning 3’ regions immediately downstream of each gene were also performed as controls for the transcription factors' binding specificity to the regulatory regions (data not shown). To correlate the ChIP results with transcriptional regulation, we also performed quantitative RT-PCR for each of the regulatory factors in the retinae of the four strains (Table 2). These results, combined with what we have learned from the literature, are discussed below.

2.1. Opposing regulation of subtype-specific genes

It is poorly understood how each photoreceptor subtype expresses the genes that determine its own identity but shuts off expression of genes specific to other subtypes. Figure 3 shows that in wild-type retinae (“WT” lanes) all five photoreceptor-specific transcription factors bind to both rod and cone target genes, regardless of subtype association. Each of the transcription factors assayed is found on the promoters of rhodopsin (Rho), cone opsins (Sop and Mop) and Rbp3, genes expressed in photoreceptors, but not the control gene GluR6 that is not expressed in photoreceptors. This suggests that each subtype-specific factor could play opposing roles on the expression of its own subtype-specific genes vs. genes specific to other photoreceptor subtypes. The best-understood example of this is Nr2e3, which is known to activate rhodopsin but repress cone opsin genes (Cheng et al., 2004; Peng et al., 2005). This is reflected in Figure 3 in the “Nr2e3” panel by the presence of Nr2e3 on all three opsin promoters, and in Table 2 by a decrease in Rhodopsin expression (0.84 times WT) and increases in S- and M-opsin expression (1.23 and 1.19 times WT, respectively) in the Nr2e3^−/− mouse compared with WT. Similarly, Nrl also directly binds to both rod and cone gene promoters in rods (Figure 3, “Nrl” panel), although a conventional Nrl target binding site has not been reported in the cone opsin regulatory sequences. This binding is independent of Crx, since it still occurs in the Crx^−/− retina. Whether Nrl binds to cone promoters through interaction with other proteins or as a result of the presence of low affinity binding site(s) in the cone promoters remains to be determined. In any case, the presence of Nrl on these promoters suggests that Nrl is involved in regulating both rod and cone genes in the same cell (Peng and Chen, 2005). Consistent with this, in the Nrl^−/− retina Rho gene expression is decreased (0.07 times WT) while cone opsin gene expression is increased (2.14 and 1.17 times WT; Table 2). Thus, Nrl exerts opposing effects on rod and cone gene expression in vivo. However, its repressive effect on cone genes is likely mediated by indirect mechanisms, as Nrl (alone or in combination with Crx) does not repress M- or S-opsin promoter activation in transiently transfected HEK293 cells (Peng et al., 2005).

The cone transcription factor Trβ2 binds to both M-opsin and S-opsin promoters in WT retinae (Figure 3 “Trβ2” panel) to activate M-opsin but repress S-opsin expression (Ng et al., 2001; Yanagi et al., 2002). Trβ2 also binds to the rhodopsin promoter (Figure 3) and several other rod genes (data not shown) in a Crx-dependent manner, suggesting that it could also be involved in regulating (likely repressing) the expression of rod genes in cone cells. Although no rod abnormalities have been reported in several different genetically engineered mice with disruptions of Thrb/Trβ2, this would be an interesting hypothesis to test. The outcome is likely to depend on interactions with other nuclear receptors and the availability of ligands and cofactors.

2.2. Auto-, para- and feedback regulation

Figure 3 also shows that each transcription factor binds to its own promoter as well as those of the other regulators examined (“WT” lanes in each panel), suggesting that each factor regulates its own expression (auto-regulation), regulates the other factors acting in parallel or downstream (para-regulation), and feeds back regulatory information to the promoters of the upstream factors that induced it (feedback regulation). The best example of this is Crx. First, the Crx protein directly binds to its own promoter [Figure 3 and (Furukawa et al., 2002)] and auto-activates its own expression. The strength of Crx activation depends on the amount of Crx present. In Crx knockout mice that have low (Crx^+/−) or no (Crx^−/−) Crx protein as a result of replacement of the Crx coding sequence of one or both alleles, respectively, transgenic LacZ reporter genes driven by Crx promoter sequences are only expressed at 57% or 31% of wild-type levels, respectively (Furukawa et al., 2002). Thus, auto-activation substantially increases expression levels. Second, Crx also binds to the promoter of other photoreceptor transcription factors Nrl, Nr2e3, Trβ2, Rxrγ, Rorβ, and NeuroD1 (Figure 3 “Crx” panel) and regulates their expression in para-regulatory fashion [Table 2; (Furukawa, 1999; Blackshaw et al., 2001)]. In Crx^−/− mice, expression of the rod factors Nrl and Nr2e3 is significantly reduced but not abolished (0.67 and 0.61 times WT; Table 2), consistent with the dramatic reduction in rhodopsin expression (0.12 times WT). Crx is also required for the expression of the M-cone factor Trβ2 by binding to the promoter of the Trβ2 gene (Figure 3). Trβ2 transcription in the Crx^−/− retina is only half (Table 2) of the wild-type level, consistent with lack of Trβ2 binding on target cone genes (Figure 3, “Trβ2” panel) and defective cone opsin transcription (Table 2) in the Crx^−/− retina. Likewise, Crx may activate the expression of NeuroD1 by binding to its promoter (Figure 3), as NeuroD1 expression is also decreased (0.75 of WT level) in Crx^−/− mice [Table 2 and (Blackshaw et al., 2001)]. NeuroD1 has been implicated in the survival and maintenance of both rods and cones (Morrow et al., 1999; Pennesi et al., 2003), and it may play this role by regulating the expression of rod/cone genes and their transcription regulators as suggested by the results shown in Figure 3. Although the reduced (0.75 times WT; Table 2) levels of NeuroD1 apparently do not dramatically affect its binding to target genes in the Crx^−/− mice (Figure 3, “NeuroD1” panel), our results suggest that insufficient NeuroD1 might contribute to the photoreceptor degeneration in these mice. Interestingly, Rxrγ and Rorβ2 mRNA levels are elevated in the Crx^−/− retina [Table 2 and (Blackshaw et al., 2001)]. This suggests that Crx plays a negative regulatory role on expression of these two factors, although the mechanism for this remains to be determined. Finally, Otx2 has been shown to act upstream of Crx by binding to and activating the Crx promoter (Nishida et al., 2003; Nishida, 2005). Crx also binds to the Otx2 promoter to repress rather than activate Otx2 expression, since Otx2 transcript levels are increased more than two fold in Crx^−/− mice (Table 2). This is consistent with the fact that Otx2 expression is down-regulated in the photoreceptor cell layer when Crx expression reaches a high level during normal retinal development (Nishida et al., 2003). Thus, at the center of the photoreceptor transcription factor network, Crx regulates expression of itself (auto-regulation), other members acting downstream or in parallel (para-regulation) and its upstream inducer (feedback regulation).

Similar auto- and para-regulation may also apply to photoreceptor subtype-specific regulatory factors. For example, the rod factor Nrl binds to its own promoter and auto-activates it, as only 40% of WT reporter transcript levels are seen in Nrl^−/− mice carrying a transgene under the control of the Nrl promoter (Yoshida et al., 2004). Besides regulating the expression of Nr2e3 in the rod pathway, Nrl also binds to the Trβ2, Rxrγ, and Rorβ promoters in the cone pathway (Figure 3). This likely results in repression of Rxrγ, and Rorβ, as they are increased in Nrl^−/− retinae [Table 2 and (Yoshida et al., 2004)]. Although we did not detect significant changes in Trβ2 expression levels or in Trβ2 protein binding to target promoters in Nrl^−/− or Nr2e3^−/− mice, under physiological conditions in the wild-type background Nrl could play a role in repressing Trβ2 expression either directly or indirectly. For the cone pathway factors, Trβ2 binds to its own and other cone and rod regulator genes, raising the possibility that it could also mediate repression of these regulators to reinforce the M-cone pathway. Trβ2 also binds to the promoters of upstream regulators Crx and Otx2, which could mediate feedback regulation of these factors. It would be interesting to evaluate this hypothesis by examining expression of these upstream regulators in Trβ2 knockout mice.

2.3. Protein-promoter interactions can be affected by protein-protein interactions and accessibility of individual promoters

The ChIP results presented in Figure 3 and expression level data in Table 2 also suggest crosstalk between protein-promoter interactions and protein-protein interactions for the photoreceptor transcription factors. As an example, Nr2e3 binding to its target genes, including both opsins and transcription regulators, appears to depend on its interacting partner Crx, as this binding does not occur in Crx^−/− mice. This lack of Nr2e3 target binding cannot be fully explained by the moderate reduction in Nr2e3 expression (60% of wild-type level) in Crx^−/− retina, as rd7 heterozygous mice that produce half the normal amount of Nr2e3 protein have normal Nr2e3 function. Furthermore, in homozygous Crx^−/− mice, Nr2e3 still binds to the Crx-independent gene Rbp3. Thus, Crx-dependent genes require either a high dose of Nr2e3 to bind to their promoters or else the presence of Crx to recruit Nr2e3 binding and regulation. Similarly, Trβ2 also appears to show such Crx-dependency in binding and regulating target genes (Figure 3, “Trβ2” panel), although no direct Trβ2/Crx interaction has been reported. In Crx^−/− mice, consistent with the reduction of Trβ2 mRNA shown in Table 2, the Trβ2 protein level is also moderately reduced on Western blots (data not shown). This reduction cannot fully account for the lack of Trβ2 binding to rod or cone target genes in Crx^−/− mice, suggesting that Trβ2 binding to target genes depends on Crx or other Crx-regulated factor(s). Since many nuclear receptor consensus binding sites in photoreceptor gene promoters appear not to favor binding of nuclear receptor homodimers, these results raised the possibility that other nuclear receptors known to regulate photoreceptor genes might also function in a similar way, i.e. interacting with Crx/Otx2 to bind and regulate photoreceptor genes. In contrast, Nrl does not appear to be required for binding of the nuclear receptors to photoreceptor genes, as Trβ2 binds to its targets well in Nrl^−/− mice (Figure 3). No Nr2e3 target binding is detected in Nrl^−/− mice because Nr2e3 is not made in this genetic background [(Mears et al., 2001); and Table 2]. We have not yet observed dramatic changes in target binding for Nrl and Crx in any of the mutant strains tested, suggesting their binding is independent of the other factors examined.

One exception for Crx-dependent binding of nuclear receptors to target genes is the Rbp3 gene. Rbp3 is known to be a target of Otx2 (Fong and Fong, 1999) and is expressed early during retina development, at the time Crx is turned on (Bibb et al., 2001). Thus, Otx2 might assist Nr2e3 and Trβ2 binding to the Rbp3 promoter in Crx^−/− mice. Another possible explanation is that the Rbp3 promoter is more accessible to regulatory factors so that a low concentration of a nuclear receptor in the presence of Otx2 is sufficient for Rbp3 promoter binding. It would be interesting to compare histone modification and other chromatin configuration markers between the Rbp3 promoter and the other promoters, to determine what makes Rbp3 more accessible to regulatory factors.

3. CONCLUSION

Taken together, the recent progress in understanding the molecular mechanisms controlling photoreceptor subtype development and the results of the combined ChIP-expression analysis presented above suggest that the photoreceptor transcription factors form two types of network: protein-protein interaction and protein-promoter interaction. The information for these two networks is just beginning to emerge. Figure 4 shows our current model for the protein-promoter interaction network. Although this model is still missing components and connections, it does offer an overview of how this protein-promoter interaction network coordinates the auto-, para- and feedback regulation among photoreceptor transcription factors that determine general or subtype-specific photoreceptor lineages. This regulatory network is essential for precisely controlling spatial and temporal photoreceptor gene expression, development and maintenance. Therefore, perturbing any of the components, either by mutations or changes in expression levels of factors, could potentially disturb the balance of the network and result in developmental defects or degeneration of particular photoreceptor subtypes. Understanding this network is important for future therapeutic interventions to treat those diseases associated with photoreceptor transcription factors.

Figure 4. Model for transcription factor network regulation of photoreceptor subtype development.

Photoreceptor subtypes develop from photoreceptor precursors derived from multi-potent progenitors via three major pathways (thick arrows). Photoreceptor transcription factors that play a major role in this process are listed based on their epistatic relationship as determined by in vivo and/or in vitro functional studies. Thin lines show protein-promoter interactions; solid lines show interactions reported here and/or previously; dotted lines are from unpublished data. Arrows indicate positive regulation, while blocked lines represent inhibition/suppression. Absence of lines indicates that the relationship remains to be determined.

4. EXPERIMENTAL PROCEDURES

Animals

All experimental procedures were pre-approved by the Institutional Animal Care and Use Committee of Washington University School of Medicine, and conformed to the guidelines of the Association for Research in Vision and Ophthalmology for the use of live animals in vision research. Mice were bred and maintained in barrier facilities at Washington University School of Medicine under a 12-hour light, 12-hour dark cycle. C57Bl/6J and rd7 mice were originally purchased from the Jackson Laboratory. Nrl^−/− mice were obtained from Anand Swaroop at the University of Michigan. Crx^−/− mice were kindly provided by Connie Cepko at Harvard Medical School.

Chromatin immunoprecipitation

The protocol used for chromatin immunoprecipitation has been published (Peng and Chen, 2005). Briefly, DNA and chromatin in pooled nuclear extracts from 6 retinae were cross-linked with formaldehyde prior to immunoprecipitation with specific antibodies. Antibodies used in the work presented here that are not referenced in Peng and Chen (Peng and Chen, 2005) include: rabbit anti-Nrl [Chemicon; see (Swain et al., 2001) for specificity details], rabbit anti-Trβ2 [Upstate; see (Srinivas et al., 2006); and (Yen et al., 1992)] and goat anti-NeuroD1 [Santa Cruz sc-1084; see (Acharya et al., 1997); and (Cissell et al., 2003)]. Antibody specificity was confirmed by Western blotting and immunohistochemistry on retinal sections, comparing WT and the appropriate knockout mouse retinae; or the use of antibodies from different sources, which gave identical results (data not shown). The results of ChIP assays were analyzed using candidate gene-based PCR with primers spanning the promoter region of each gene (listed in Peng et al., 2005 or shown in Table 3). PCR assays with primers spanning 3’ regions immediately downstream of each gene were also performed as controls for factors' binding specifically in the regulatory regions (Peng and Chen, 2005). Results shown are representative of at least three separate experiments. Controls include the use of normal rabbit/goat IgG (Santa Cruz) in immunoprecipitation reactions (negative controls) and input (without ip) as positive controls in PCR reactions.

Table 3.

PCR primers for ChIP and quantitative RT-PCR

Genes	Assays	Sense	Anti-sense
Trβ2:	ChIP (−321/−53)1	5′-ACCTGCCTGCCATTTTCCC-3′	5′-ATTTGCCAGCCCCCTGAAC-3′
	qRT-PCR2	5′-GCACATCTCCCTGAAGAAAAGC-3′	5′-TCCCCACACACTACACAGAGC-3′
NeuroD1:	ChIP (−1504/−1336)	5′-TCCAGCCACTCAACCCTGAC-3′	5′-GAGGAGGAGGAGGAATGGTG-3′
	qRT-PCR	5′-CGCTCAGCATCAGCAACTC-3′	5′-CTTGTCTGCCTCGTGTTCC-3′
Otx2:	qRT-PCR	5′-ACTTGCCAGAATCCAGGGTG-3′	5′-TGAGCCAGCATAGCCTTGAC-3′
Crx:	qRT-PCR	5′-TGTCCCATACTCAAGTGCCC-3′	5′-TGCTGTTTCTGCTGCTGTCG-3′
Nrl:	qRT-PCR	5′-TTCTGGTTCTGACAGTGACTACG-3′	5′-AAGGCTCCCGCTTTATTTC-3′
Nr2e3:	qRT-PCR	5′-AGTCCCAGGTGATGCTAAGC-3′	5′-TTCTAAGATGTGCTGCCCC-3′
Rxrγ:	ChIP (−362/−109)	5′-AAAGGGCTCTGTTCTCTCTTGG-3′	5′-CGGGTGGCACAATCTATTAGC-3′
	qRT-PCR	5′-CAATGCTCTTGGCTCTCCG-3′	5′-ATCTTTGTTATCCCGACAGGTG-3′
Rorβ:	ChIP (−471/−96)	5′-AAAGAGACAGAGGAGAGAGGGG-3′	5′-CAGTTAGAGGATGCTGGGTGC-3′
	qRT-PCR2	5′-AAGGGATTCTTCAGGAGGAGC-3′	5′-CCGCTGCTTCTTGGACATC-3′

¹For ChIP primers, the numbers in parentheses indicate the position relative to the transcription start site as +1.

²Note that qRT-PCR primers for Trβ2 recognize only the β2 isoform, while Rorβ primers recognize both β1 and β2 isoforms.

Quantitative real-time PCR

The protocol used for quantitative RT-PCR has been published (Peng and Chen, 2005). Sequences for additional primers used in this study are shown in Table 3. Briefly, cDNA reverse-transcribed from 1ug total RNA was diluted 10-fold and quantified by real-time PCR analysis in triplicate on an iCycler PCR machine (Bio-Rad), using SYBR Green JumpStart ReadyMix (Sigma). β-Actin was used as a loading control. Relative expression levels were normalized to the β-Actin levels for each sample according to standard methodology (http://www.openlink.org/dorak), as follows: ΔCT = CT_(test) – CT_(β-actin) where CT, the threshold cycle, is the cycle number (in the exponential phase) at which enough amplified product has accumulated to yield a detectable fluorescent signal that is significantly above the baseline fluorescence level. Results are presented as the ratio of ΔCT_knockout / ΔCT_WT. Mean values and standard deviation (STDEV) were calculated for each experiment from three replicates, and statistical significance was determined using the paired student t-test.